Learning residual refinement network with semantic context representation for real-time saliency object detection

作者：

Highlights：

• We design a GAM to generate a global semantic context guided representation, which is transmitted into all the bottom layers.

• We develop an MRB that aggregates the features with multi-scale contextual information and enlarged receptive fields.

• We build an RRM to learn the residual that focuses on the boundary details recurrently. Our SOD algorithm achieves leading performance on 6 benchmark datasets compared to the state-of-the-arts with a speed of 29 fps.

摘要

•We design a GAM to generate a global semantic context guided representation, which is transmitted into all the bottom layers.•We develop an MRB that aggregates the features with multi-scale contextual information and enlarged receptive fields.•We build an RRM to learn the residual that focuses on the boundary details recurrently. Our SOD algorithm achieves leading performance on 6 benchmark datasets compared to the state-of-the-arts with a speed of 29 fps.

论文关键词：Salient object detection,Convolutional neural networks,Deep learning,Residual learning

论文评审过程：Received 15 July 2019, Revised 9 February 2020, Accepted 10 April 2020, Available online 19 April 2020, Version of Record 5 June 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107372